ARTIFICIAL INTELLIGENCE IN COLONOSCOPY: EVALUATION OF ADENOMA DETECTION RATE AND PERFORMANCE CHARACTERIZATION

Area of Science:

• Gastroenterology
• Medical Technology
• Oncology

Background:

• Colonoscopy is crucial for detecting colorectal cancer precursors.
• Artificial intelligence (AI) shows promise in improving lesion detection during colonoscopy.

Purpose of the Study:

• To compare the adenoma detection rate (ADR) in colonoscopies performed with and without AI software.

Main Methods:

• An observational study included 305 patients undergoing screening colonoscopy at a tertiary hospital in Brazil.
• Patients were randomly assigned to conventional colonoscopy (CC) or AI-assisted colonoscopy (Cadeye system - CADe).
• Data on patient, procedure, and polyp characteristics were collected.

Main Results:

• A total of 470 polyps were detected across 305 colonoscopies (112 CADe, 193 CC).
• Overall ADR was 53.8% and overall polyp detection was 74.8%.
• No statistically significant difference in ADR was observed between CC (57%) and CADe (48.2%), with P=0.138.

Conclusions:

• While some studies suggest AI increases ADR, this study found no difference between AI and conventional colonoscopy.
• The added benefit of AI in colonoscopy may be limited in settings with a high baseline ADR.
• Further research may be needed to clarify AI's role in colorectal cancer screening.